Multiple hypotheses data association propagation for robust monocular-based SLAM algorithms

Data Association is probably the most important step of every monocular Simultaneous Localization and Mapping (SLAM) algorithm because it provides the basic information to the estimation module, independently on the estimation algorithm of choice. Although important, it is also a difficult task because the analytic solution is NP-Hard. The usual approximation is obtaining only one data association hypothesis per frame which affects the robustness of the result [1][2][3][4][5]. In this paper, a data association approach is presented, where multiple hypotheses are propagated between frames using a probabilistic framework. Experimental results, using real and synthetic data, show that the proposed algorithm produces promising results with respect to other state of the art methods.